Improve Text Classification Accuracy by 10% to 15%
$10-30 USD
Pago na entrega
I have an SQL database schema with thousands of events (30,000 to be exact but only about 5,500 are currently usable). What I'd like to do is improve my current 70% f1-score by at least 10% (I can get nearly 90% success with the 30,000, but not with the 5,500). I am currently using Scikit-Learn's machine learning library. It's written in python.
What I'd prefer is someone who could simply apply better feature selection. I've implemented the Chi-squared and best-features options within Scikit, but I've had limited success. I'd like someone who could explain how to implement a high-value terms selection criterion. Something along these lines: [url removed, login to view]
You don't need to write the code yourself -- tho if you're interested I'll provide the database schema, so you can apply a library of your choice (preferably Scikit or Solr). I need someone who can help explain the process. I can write everything. I'm just not an ML expert. Thank you in advance for any interest.
ID do Projeto: #5269629
Sobre o projeto
Concedido a:
My suggestion (assuming you have a very large dictionary/feature space): - Do TFIDF (if you haven't done done already) - Do hashing on features to reduce the size of your feature space (this works surprisingly well i Mais
5 freelancers estão ofertando em média $47 nesse trabalho
I have previous experience in ,achine learning development, I would like to know more about the project....